22 comments

[ 3.2 ms ] story [ 55.6 ms ] thread
This company is making all the "smart" tech companies in the bay area look foolish
Now they'll pay for training data? Wonder if this will impact lawsuits against them for previously used training data.
I don't think that is what reported here, they just say that if you have a lot of data and you want the model to be better at that domain, you can collaborate with OpenAI because both parties are interested.
Basically they take your data for free and they will receive your money from the API usage you will make.

Good

isn't that literally every software product? You provide the data and they provide value on top
If we’re all accepting that private training data is valuable, there is no valid argument that OpenAI should have free access to copyrighted material for training data without permission.
Sure there's an argument:

Information wants to be free. Copyright is abusive. Broad fair use rights, especially for transformative purposes, is all-important for the future.

The costs of enforcing limits are higher than the benefits of letting AIs have an equivalent 'right to read' as humans.

Not everything that's 'valuable' can or should be 'owned' & locked behind fidgety statutory permissions.

OpenAPI’s own data is closed and… wait for it…copyrighted.

So it’s ok for them to steal other folks work as long as no one can steal theirs.

Cool story tech bro.

It's murky whether the outputs of OpenAI's models can be copyrighted at all.

Further, even the model weights, if released or leaked, might not qualify.

Even 'steal' isn't quite the right word for the allegation of the malum prohibitum 'crime' of 'intellectual property infringement'. That's especially so if the disliked use isn't a reproduction as described in current law.

You might have a better case trying to extend the fuzzier concept of 'moral rights'. The distaste for the training, without it crossing the specific lines in copyright law, more closely resembles that vague form of IP.

But hey, you have in fact offered a sort of emotionally-rooted argument, albeit unconvincing, rather than the grandparent's unargued mere conclusory opinion that no valid contrary arguments exist.

Under current copyright law, I do not have a right to own a copy of a copyrighted work that I obtained without the permission of its owner. “Information wants to be free” is a fuzzy assertion of ideology that does not have any basis in intellectual property law. “Fair use” is an argument about published works, not the right to possess the underlying copyrighted work in the first place.
"We are not seeking datasets with sensitive or personal information, or information that belongs to a third party; we can work with you to remove this information if you need help."

One thing that is impressive about OpenAI, apart from the technology, is that their written materials are really good. This sentence in particular is very artfully done.

Huh? It's a rather basic sentence? Is the use of a semicolon considered art these days?
Yes: basic but thorough in assuaging many concerns, in a concise quotable way.
Let's be real. At some point someone (a government like China) will train a model on all available copyrighted materials and will have a huge advantage in this field. Something needs to be done. Not sure if a private company like OpenAI should be one who's doing it though
poster-child commentary rationalization of anything by conjuring demons.. call the exorcist!
And that's not counting the data advantage China already has with the large scale surveillance it does.

US regulators act like they control the world

That sounds awesome. I can’t wait.
I think Japan even passed a law allowing this.
This seems like a lot of words to just say "FEED ME!"
I assume this is unopenai trying to build themselves a moat?